Evolution of title of long ACL papers. Each column shows terms distinctive to a set of consecutive years. Size indicates frequency; order characteristicness.

1979-1983	1984-1988	1989-2008	2009-2010	2011-2015	2016-2018	2019-2022
natural language interface	project	machine learning	web search	joint inference	attention model	bert
design	tree adjoining grammars	unification - base	expectation	distributional semantics	lstm	pre - train
issue	machine translation system	automatic acquisition	letter	segmentation use	neural semantic	pre - training
access	computer	lr	phrase - base statistical machine	dual decomposition	deep reinforcement learning	transformers
production	sharing	maximum entropy	system combination	decipherment	sequence learning	few - shot
database	design	lexicalized	hierarchical phrase - base translation	acl	skip	multi - hop
computer	natural language interface	corpus - base	binarization	reorder	sequence models	pretraine
speech act	production	large corpora	base statistical machine translation	recursive neural	deep neural networks	pretrained
interface	unification grammar	co - occurrence	phrase - base translation	unsupervised semantic	bandit	pre - train language model
experimental	categorial	unification	semi - supervised learning	walk	character - level	contrastive learning